Purpose
To reproducibly montage ultra-widefield (UWF) fundus images and accurately quantify retinal surface area in mm2, utilizing the Optos 200Tx(Optos, UK).
Methods
Montages were created from five fundus photographs taken by the Optos 200Tx at directionally-guided gaze angles. Every image was transformed into a stereographic projection of the eye. This transformation first mapped each pixel to a 3D model eye with a 24mm diameter and then to a stereographic projection by projecting all pixels to a plane through the equator of the eye. This was performed by ray tracing pixels in the image through a combined optical model eye. For each image, the location of the fovea was used to guide the mapping table for the corresponding gaze angle. Then, the four peripherally eye-steered stereographic images were registered to the on-axis stereographic image. Image registration extracted the vasculature in image pairs and cross-correlated. The pixels to be quantified were mapped from the montaged 2D stereographic projection to the 3D spherical representation. Spherical trigonometry was then used to quantify area.
Results
To assess precision, 10 fluorescein angiograms (FA) of the same healthy eye were performed. A montage was produced from each of these and 4 vascular landmarks were selected, one in each quadrant. A line was drawn between each of these landmarks to create a standardized quadrilateral. The mean area of the quadrilateral was then quantified across the 10 FA montages, with a high degree of concordance (mean=408.97mm2, range=8.54 mm2, SD=2.91mm2, RSD=0.7%, variance=8.49 mm2). To assess accuracy, the mean optic nerve head disc-area (DA) of 50 patients was quantified and compared to a reported meta-analysis (n=9852; mean=2.15mm2, SE=0.06mm2). The mean DA was 2.21mm2 (SE=0.06mm2), a difference of 2.7%. To assess distortion inherent to a montage, 1 DA was defined by the circle of pixels comprising the optic disc in a given eye (7,667px, 2.35mm2). This circle of pixels was translated radially and quantified every 50px until reaching the ora serrata. The circle represented 1.39mm2 and 0.32mm2 after being translated by 1000px and 2000px respectively, a difference of 41% and 86% from actual size (FIGURE AB).
Conclusions
A high degree of accuracy and precision in montaging and quantification was achieved. With an effective method to quantify area from UWF images, many clinical and research applications may be realized.
Keywords: 549 image processing •
550 imaging/image analysis: clinical •
551 imaging/image analysis: non-clinical